A mobile application for robust feature extraction and cultivar classification of leaves

We illustrate the development of an application for cultivar classification of leaf images based on the extraction of the network of its main veins that runs on mobile devices like smart phones or tablets. Such mobile devices can be docked to farming robots in order to support the farming process. Our application uses an efficient Gabor filter-based tracing algorithm which is able to perform a robust network extraction. The results are used as input data for the classification with a support vector machine. In order to demonstrate the advantageous behavior and the robustness of this method, we perform an evaluation on a test set consisting of 150 light transmitted images of different vine leaves.